proton treatment planning

Proton Treatment Planning

November, 9th, 2012

AAPM GLCM, Flint, MI

Stefan Both, PhD

2222

Outline

� Proton Therapy @ UPenn

� Principles of Proton Therapy and Treatment Planning

� PBS Clinical Implementation: Penn Solutions &future work

� Summary

3333

4444

4

Description of the Roberts Proton Therapy Center

• 4 gantries + 1 fixed-beam room + 1 research room

• 2 gantries have universal nozzles with SS, DS, US, PBS &

MLCs

• 2 gantries have universal nozzles with SS, DS, US & MLCs

• Fixed-Beam-Room has dedicated PBS nozzle

• All patients are setup with orthogonal x-ray (G=270

degrees)

• All gantries have MLCs with two compensator mounts

6666

6

Delivery Methods- Passive Scattering

� Accelerated protons are near monoenergetic and form a beam of small lateral dimension and angular divergence

� Single Bragg Peak spread out by range modulator

� Field Profile spread laterally by a set of spreaders compensated for the range

� Beam Shaping:

-Block/MLC Laterally and Compensator in Range(Distally)

7777

7

Delivery Methods: Pencil Beam Scanning

A PB is scanned both laterally and in depth ( by changing its energy) => in a near arbitrary dose distribution laterally and dose sharpening in depth (Pedroni et al.)

- lateral distribution determined by the lateral positions and weights of each pencil beam of a chosen energy- Isolayers

- distribution in depth is determined by weighting the pencil beam at each position within the field.

8888

A proton pencil

Beam Spot3...

A few pencil beams

together3.Some more3

A full set, with a

homogenous dose

conformed distally and

proximally

Pencil-Beam Scanning – PBS

Images courtesy of Eros Pedroni, PSI

Magnetically scan p beam left / right (X,Y) and control

depth with Energy (Z)

Fully electronic and no mechanical parts!

9999

Relevance of pRT

�PT and XRT treatment history is inversely symmetric

• Emphasis of XRT was to increase conformality – IMRT

• Emphasis of PT must be on PBS and promulgate

Courtesy of Hanne Kooy

�p always has “superior” dose distributions

�; but does not treat enough sites

• Not Quantitatively (< 1%)

• Not Qualitatively (prostate)

Co

nfo

rma

lity

TimeNow

ProtonsPBS

IMRT

3D

Nu

mb

er

of

site

s

TimeNow

?

100

10

1

10101010

Principals of PROTON Therapy and Planning

Contrast with photons (x-rays)

- Photons continue to deposit dose beyond target in

tissue3.

..while normal tissue radiation offers no advantages for the patient

11111111

RBE and OER for Protons

12121212

Physics of p is understood;

Proton beam could be shaped and manipulated completely

by mechanical means- passive scattering, >50 yrs.

�Passage through an absorber means

• Reduction in energy but NOT intensity (number)

• Dispersion (scatter) of beam

Absorber

13131313

13

Tracks in PatientCourtesy of Hanne Kooy

mp= 2,000 me

14141414

Normal Tissue Exposure to Radiation Dose

15151515

15

Planning of Proton Therapy

� Illustration of the volume and margins relating to the definition of the

target volume per ICRU 62:

16161616

16

Planning of Proton Therapy

� Volumes and margins related

to the OARs:

17171717

17

Planning of Proton Therapy

Proton –specific issues related to the PTV

� For photon beam the PTV is primarily used to delineate the lateral margin

� For protons in addition to lateral margins a margin in depth has to be left to allow for uncertainties in the knowledge where the distal 90% IDL would fall

� Proton Beam Energy should be selected in a way that the CTV is within the irradiated volume taking into account both motion and range uncertainties

� Since the lateral and the margins in depth solve different problems each beam orientation would need a different PTV

� Alternatively the beam parameters are determined based on the CTV adding the lateral and range margins to the TPS alg.

18181818

18

Planning of Proton Therapy

� In practice the beam parameters are determined based on

the CTV adding the lateral and range margins to the TPS alg

for each beam.

� For scanned Beams and IMPT these margins would influence

which pencil beam would be used and each one’s depth of

penetration. It is much easier to visualize using optimization

volumes( PBSTV)

� It is “required” that the dose distribution within the PTV is

recorded and reported , therefore a PTV relative to CTV

based on lateral uncertainties alone is proposed by ICRU 78

� We can safely do this is we ensure plan robustness first.

19191919

Planning of Proton Therapy

Sources of uncertainties:

� Patient related: Setup, movements, organ motion, body contour, target definition, etc3

� Physics related: CT number conversion, dose calculation, etc3

� Machine related: Device tolerances, beam energy, delivery method, etc3

� Biology related : Relative biological effectiveness ( RBE), etc..

20202020

Uncertainties in Proton Therapy

“If something goes wrong in the planning process it starts usually at the CT Simulator ;”

Physics Issues:

� CT Calibration Curve:

- Proton interaction ≠≠≠≠Photon interaction

- Multisegmental curves are in use

- No unique SP values for soft tissue HU range

- Tissue substitutes ≠≠≠≠ real tissues

- Statistical and systematic variations in CT numbers

- Image reconstruction artifacts ( High Z materials)

21212121

Uncertainties in Proton Therapy CT Calibration Curve Stoichiometric Method

22222222

22

Uncertainties in Proton Therapy CT Calibration Curve Stoichiometric Method

Is the 3.5% CT# correction for proton range uncertainty conservative?

Experimental evaluation of the relationship between the CT#

and proton stopping power ratio was done at PSI using a stoichiometric method ( Schaffner et al 1998, PMB)

Conclusion: There is a 1.1 % uncertainty in soft tissue and 1.8% in bone.

Reality;A decade later it is still NOT the current clinical practice !

3.5% standard;

23232323

Uncertainties in Proton Therapy CT High Z artifacts

� Artifacts due to high Z materials (metal clips, fiducials, Calypso

beacons, prosthesis, dental fillings, etc.) are common in RT.

� Avoid beam paths through high Z structures.

� Range uncertantanties in proton therapy due to significant CT

reconstruction artifacts require to increase the typical 3.5%

range uncertainty to 5% for the distal margin after manual

clean up of the CT image by the planner.

24242424

Uncertainties in Proton Therapy CTHigh Z artifacts

Note: Image quality improvement for diagnostic purpose do not account for HU corrections at an accuracy level required for calculations in RT

25252525

Proton Treatment Planning: Inhomogeneitis

� The effect of tissue inhomogeneity

is greater for protons then for photons

(ICRU 78)

� Failure to allow for a higher density

along the proton path may result in a

near zero dose in a distal segment of

the target due to the reduced range

of the protons.

� Penumbra is minimally affected for the

materials limited to the human body, but

it changes significantly for other material

as it is caused by multiple scattering

� Conversely neglecting to account

for an air cavity upstream of the target

=> in high dose deposited in distal

normal structures.

26262626

Uncertainties in Proton TherapyMotion and Setup uncertainties

� What happens if the beam is nearly tangential to the target?

� Therefore, tangentials fields are avoided in clinical practice

ICRU 78

27272727

Planning of Proton Therapy

RBE Uncertainties

� Clinical RBE: 1 Gy proton dose ≡ 1.1 Gy Cobalt γ dose (RBE = 1.1 in the middle of SOBP)

� RBE weighted dose concept introduced by ICRU 78

� RBE vs. depth (LET) is not constant

� RBE also depends on

• dose

• biological system (cell type)

• clinical endpoint (early response, late effect)

� How do we overcome this uncertainty in clinical practice?

In general, not more then 2/3 of our prescribed dose comes from beams pointed towards a critical structure.

28282828

PBS Planning Techniques

� PBS based treatment planning can be performed using two

different techniques:

� Single field optimization (SFO)- where single fields are

optimized to achieve uniform dose (as known as SFUD).

� Multifield optimization (MFO, IMPT)- where all spots from all

fields are optimized simultaneously, and dose in each single

field is not uniform (similar to IMRT).

29292929

SFO (SFUD) vs. MFO

SFO RT

MFO LT

LT

RT

30303030

Optimization Volume-PBSTV

� Beam specific PTV margins are related to the range uncertainties and

incorpoated in the optimization volume-PBSTV.

Distal and proximal margins are set from CTV:

• DM = (0.035 x CTVdistal) + 1 mm

• PM ≈ (0.035 x CTVproximal) + 1mm

- Lateral margins based on setup, motion, penumbra.

3.5%- uncertainty in the CT# and their conversion to relative proton linear stopping power

1 mm - added to correct for range uncertainty

31313131

SFUD vs. MFO vs. Passive Scattering

� Double scattering for moving targets

� Uniform scanning for sharp penumbra, larger field, deep

seated tumor

� SFUD for highly conformal dose distribution

� MFO is currently not employed at Penn

Conformality Robustness Planning

Best Worst Best Worst Easiest Hardest

MFO SFUD PS PS SFUD MFO MFO SFUD PS

32323232

Clinical Implementation:Base of Skull RT

Some tumors require high dose of radiation (> 70Gy)

while we have:

� Limited dose level tolerances for brainstem, optical

chiasm, optical nerves, cochlea , etc..

� To decrease the amount of normal brain irradiated

With PBS:

� Rapid dose fall off achievable through small pencil beam

size

� Proximal and distal dose conformality

� Reduced integral dose

33333333

Range Shifter & Spot Size

� The fix beamline has energy range (100 MeV to 235 MeV)

� For targets <7cm from the surface require the use of energy

absorber (range shifter)

� Range shifter positioned at

the surface of the snout with

>30cm air gap to ISO

100 120 140 160 180 200

5

10

15

20

25

30

35

40

Beam Energy (MeV)

Sp

ot

siz

e (

mm

)

X w/o RS

X with RS

Y w/o RS

Y with RS

� Pencil beam spot size

increases significantly with

air gap

34343434

Bolus for Brain Tumor

� Maintain the size of the pencil beam

� Minimizing the air gap and the amount of material in the beam

� Range shifter (RS) was replaced with an Universal Patient

Bolus

120 140 160 180 200

5

10

15

20

25

30

35

Beam Energy (MeV)

Sp

ot

siz

e (

mm

)X direction

No RS

2cm Bolus

8cm Bolus

RS

120 140 160 180 200

5

10

15

20

25

30

35

Beam Energy (MeV)

Y direction

No RS

2cm Bolus

8cm Bolus

RS

35353535

Pencil Beam Scanning Technologies Spot Size Integrity - Penn Solution In Room Implementation

36363636

Clinic Example

� Target is close to brainstem, cord, cochlea and optical structures.

37373737

Pencil Beam Scanning Technologies Eclipse Bolus vs. Range Shifter

38383838

DVH Comparison: Bolus ( ) vs. RS ( )

� More uniform target coverage and superior conformality

� The biggest differences in dose for the OARs are for the peripheral

structures such as the cord and cochlea

� The brainstem and chiasm are similar in the high dose region

39393939

Choosing Beam Orientation

� Beam orientation is chosen to have the shortest and the most

homogenous distance to the target (for robustness)

� Multiple beams are used for robustness, but less beams than

DS due to TPS limitation

� Multiple beams without skin overlap to reduce the skin dose

� Avoid beams point towards critical structure due to range

uncertainty

40404040

Penn Collision Detection Software

� CAD /MATLAB ray casting algorithm.

� Incorporated during the proton treatment planning phase, to improve

clinical efficiency.

� The method could apply to patient collision detection in XRT.

Figures 3 & 4 illustrating the collision detection method (green – body

contour points; red – gantry polygon).

W. Zou, S.Both. Et al.“A Clinically Feasible Collision Detection Method for Proton Therapy” (accepted Med Phys J.).

A

B

Figure 3 Figure 4

41414141

SFUD planning in Eclipse (1) - Volume

� PBS plan needs a volume for selection of spot position

� Volume for optimization: pencil beam scanning target

volume (PBSTV) that includes range uncertainty in

beam direction

� For brain tumors, PBSTV=CTV+5mm

� Eclipse limitation: it could not add late margins in beam

direction for PBS optimization

Eclipse LimitationPBS DS

42424242

SFUD planning in Eclipse (2) - Artifacts

� All CT artifacts need to be contoured and overwritten with

appropriate HU (e.g. high density clips, BB, bone artifacts).

� It will needs to change window and level to identify them.

Clips (HU>3000)Bone artifactsBB

43434343

SFUD planning in Eclipse (3)

44444444

SFUD planning in Eclipse (4)

� Lateral margins (1-2 spot spacing) are used for extension of

dose grid, so spots can deposit outside the PBSTV in order to

achieve good coverage

Lateral margin

45454545

SFUD planning in Eclipse (5)

� Simultaneous spot optimization (without OAR constraints)

Varian

46464646

SFUD planning in Eclipse (6)

� OAR optimization (field by field)

Varian

47474747

Minimum MU

� A minimum signal-to-noise ratio is required for reliable spot

position measurement

� The spot does should be greater than the expected delayed

dose (the dose delivered after the beam spot termination

signal is sent by the main dose monitor)

� Our minimum MU is 0.021MU, ~ 60 pC

� Spot post processing

• Rounding down: spot is deleted if MU < 0.5 MUmin

• Rounding up: spot is rounded to MUmin if 0.5 MUmin ≤ MU < MUmin

48484848

Spot Post Processing� Post-processing runs automatically after the optimization and

before dose calculation

� Optimal spot weights (raw) changed after post processing

49494949

Rounding Errors – TPS Limitation

� Since Eclipse does not incorporate minimum MU constraint in

optimization, the ideally optimized dose distribution was

distorted after post processing due to minimum MU.

� The dose distribution is more distorted the plans with multiple

fields because MU for each spot is reduced.

� Do not use too many fields due to this limitation in TPS.

50505050

TPS limitation on PBS optimization

� A BOS case with four equally weighted fields.

� For this specific layer almost half of the spots were deleted

after post processing.

N=108 N=56

� TPS should incorporate MU constraints in the optimization

process!

51515151

Patient Specific QA

� Geometry: center of SOBP align with ISO, sub mm accuracy of

alignment was achieved with IGRT

� Dose maps in four depths were measured

� Absolute point dose comparisons and gamma analysis for 2D

dose map

52525252

Small Fields Dose Discrepancies

� Measured output for some brain fields (small field and lower

energy) could be 10% less than the Eclipse calculation

� Renormalization is made in TPS, and redo QA at center SOBP

Original measurementEclipse calculationRenormalized

measurement

Need times 1.1 for RBE

53535353

Renormalization - Caveat

� More spot s may appear after renormalization because more

spots may be rounded up

N=56 N=66

� Renormalized plan ≠ approved plan

� Need to remove additional spots to keep plan integrity

� QA should be performed again for center of SOBP plane

54545454

Why Small Field Need Renormalization

� Halo is produced from beam profile monitor in the upstream,

which affects more for the low energy beam (e.g. brain cases).

� Halo dose is small, but its FWHM can be more than 10cm.

� With >1000 spots in PBS field, even a low dose tail (0.1%) could

accumulate to a significant dose contribution

IC20

cm

� Primary

Gaussian

σ1=1cm,

secondary

Gaussian

(halo)

σ2=5cm.

55555555

Field Size Factor

� With one Gaussian fit for in air profile, output calculated by

Eclipse is almost a constant for all field sizes.

� Output was matched to field size about 10cmx10cm, which is an

overestimation for small fields (e.g. brain fields).

� In air measurement

of output varies with

field size

56565656

PBS Treatment Planning-ProstateInterplay Effect & Prostate Motion

� PBS delivers a plan spots by spots; layers by layers.

� Each layer is delivered almost instantaneously.

� The switch (beam energy tuning) between layers takes about 7s.

� Prostate motion during beam energy tuning causes an interplay

effect.

57575757

Pencil Beam Scanning Technologies Calypso Based SI & AP Prostate Motion For One Patient

Best

scenarioIntermediate

scenario

Worst

scenario

Both, et. al. IJROBP, 12/2011

% Time

58585858

Pencil Beam Scanning Technologies Prostate Drifting and Beam on Time (Calypso) Worst Case Scenario Patient

Beam on time of Left Lateral Field Beam on time of Right Lateral Field

59595959

Pencil Beam Scanning Technologies DVH of SFUD Plan Worst Case Scenario Patient

LT + RT LT RT

60606060

Pencil Beam Scanning Technologies Interplay Effect on Dose Distribution Worst Case Scenario Patient – Worst Fraction

Both S. Proton Treatment Planning, AAPM 2012.

Tang et al. Interplay Effect and Prostate PBS Dose Distribution (Manuscript in progress).

61616161

Pencil Beam Scanning Motion management and Tx Delivery: Is Calypso an option?

Max. dose deficit occurring within the PTV from Calypso in a proton beam as a function of the WED

from the distal PTV boundary for 3 different beacons orientations with respect to the beam direction.

Dolney D. et al. “Dose Perturbations by Electromagnetic Transponders in the Proton Environment”

(submitted manuscript).

62626262

Pencil Beam Scanning Technologies Motion management and Tx Delivery: Calypso

� If a transponder is implanted or migrates to within 5 mm of the PTV

boundary, our findings indicate the possibility for greater than 10% dose

shadow downstream of the transponder.

� Plan design with multiple beam angles to distribute the shadow over a

larger volume, or possibly increasing the dose in the expected shadow

region to offset the deficit could work.

� Electromagnetic transponders could be used for patient setup and

motion management for proton therapy provided some guidelines

regarding their placement and orientation with respect to the beam can

be met.

63636363

Proton Treatment Planning & Delivery IssuesSummary

� Uncertainties have a significant impact on dose distributions actually delivered and may affect outcome

� It is KEY to educate ourselves about the impact of uncertainties and how we account for them in planning process

� Proton RT is very different from Photon RT, as Proton RT

requires site dependent implementation.

� Once we solve the problems related to PBS deployment, it may

lead to better outcome in RT.

64646464

Thank You

Acknowledgements:

Penn Radiation Oncology